Meta's Q3 2025 Earnings Call: Contradictions Emerge on AI Infrastructure ROI, Model Integration, and CapEx

Date of Call: October 29, 2025

Financials Results

• Revenue: $51.2B, up 26% year-over-year (25% on a constant currency basis); Family of Apps revenue $50.8B, up 26% YOY; Family of Apps ad revenue $50.1B, up 26% (25% cc)
• EPS: $1.05 per share (net income $2.7B); $7.25 per share excluding one-time deferred tax charge (would have been net income $18.6B excluding charge)
• Operating Margin: 40% (operating income $20.5B)

Guidance:

• Q4 2025 revenue expected to be $56B to $59B (foreign currency ~1% tailwind to YOY revenue)
• Expect continued strong ad revenue growth partially offset by lower YOY Reality Labs revenue (lapping Quest 3S and Q3 retail shipments)
• Q4 2025 tax rate expected 12%–15%
• Full year 2025 total expenses expected $116B–$118B (22%–24% YOY); CapEx $70B–$72B
• 2026: anticipate notably larger CapEx growth and materially faster expense growth driven by infrastructure/cloud and employee compensation; budget planning ongoing

Business Commentary:

• User Engagement and Product Growth:
• Meta Platforms reported 3.5 billion people using at least one of their apps daily, with Instagram reaching 3 billion monthly active users.

• Growth is attributed to improvements in AI recommendation systems, which increased time spent on Facebook by 5% and on Threads by 10%.

• Advertising Revenue and AI Advancements:

• Family of apps ad revenue was $50.1 billion, up 26% year-over-year.

• The increase was driven by advances in AI ranking systems, which improved ad performance and efficiency.

• Capital Expenditure and Infrastructure Investment:

• Meta's capital expenditures were $19.4 billion in Q3, primarily for servers, data centers, and network infrastructure.

• This investment is part of a strategic effort to build leading AI infrastructure, supporting both core business growth and new product development.

• New Product Launches and AI Integration:

• Meta launched Vibes, a new AI creation tool, which is showing promising retention and growing quickly.
• Integration of AI models into products like Meta Superintelligence Labs aims to drive innovation and enhance user experiences across platforms.

Sentiment Analysis:

Overall Tone: Positive

• Management: "We had another strong quarter" and highlighted product momentum (3.5B daily users, Instagram 3B MAU). CFO: "Q3 total revenue was $51.2 billion, up 26%" and operating margin was 40%. CEO emphasized aggressive compute investment to scale AI and product pipeline, framing investments as high‑upside for core ads and new products.

Q&A:

• Question from Brian Nowak (Morgan Stanley): Can you quantify early AB-test signals that give confidence we'll get ROIC from the increased 2026 CapEx for models/ad ranking and how large is the Reality Labs revenue headwind in Q4 guidance?
  Response: Value-weighted conversion rates are growing faster than impressions, supporting ROI expectations for AI and infrastructure investments; Q4 Reality Labs headwind not quantified but driven primarily by lapping Quest 3S and prior retail channel timing, partially offset by AI glasses growth.

• Question from Douglas Anmuth (JPMorgan): How are you triangulating CapEx dollar growth and faster expense growth next year with core growth, earnings and free cash flow; any cash/net-cash targets?
  Response: Budgeting for 2026 is ongoing; priority is securing compute capacity for AI which likely raises CapEx/expenses, and no specific cash or net-cash targets were provided.

• Question from Eric Sheridan (Goldman Sachs): What signals do you see today on consumer interaction with Meta AI and how will scaling/exiting MSL models change Meta AI utility/behavior?
  Response: Meta AI already has >1B monthly users and usage increases as model quality improves; plugging frontier MSL models into products should materially expand utility, engagement and new product opportunities.

• Question from Mark Shmulik (Bernstein): How should we think about the scale of ads/engagement improvements vs the last two years and timing for frontier model launches or product rollouts like Vibes?
  Response: Core gains come from transferring large-model knowledge to efficient runtime models and advancing inference architectures; even small percent lifts scale to material revenue, and no specific frontier model timing was announced.

• Question from Justin Post (BofA): How are you thinking about user tools coming from AI and margins in this new content cycle versus prior cycles?
  Response: It's too early to define margins for new AI products; focus is on building high-value products and maximizing long-term profitability rather than pre-defining margins.

• Question from Ross Sandler (Barclays): Is MSL organized toward AGI/long-term goals versus immediately revenue-generating products, and is personal AI for billions still the plan?
  Response: MSL is building foundational capabilities that will be integrated across consumer and business products; the goal remains to develop novel, scalable capabilities (including personal AI) that can be productized broadly.

• Question from Mark Stephen Mahaney (Evercore ISI): What adoption signals for Meta AI are most encouraging and where are you on the roll-out-to-monetization path?
  Response: Encouraging signals: large user base and clear correlation between model improvements and increased usage; next step is integrating frontier models into products to drive further engagement and monetization.

• Question from Ronald Josey (Citigroup): Can you comment on advertiser adoption rates of end-to-end automation and how ranking/model changes are driving advertiser ROI?
  Response: Advantage+ adoption is expanding and delivers measurable ROI (e.g., 14% lower cost per lead for lead campaigns); automated solutions now run at an annual revenue run rate of ~$60B with significant room to grow adoption.

• Question from Youssef Squali (Truist): On wearables, can hardware sales recoup investment or will services/ads/commerce be required; and how will Blue Owl JV and similar financing affect CapEx recognition and future funding?
  Response: Wearables combine device sales with growing AI services and are expected to be profitable long-term if scaled; the Blue Owl JV shifts most data-center construction off-balance (Meta retains ~20% funding obligation recorded as other investing cash flows), reducing near-term CapEx recognition.

• Question from Kenneth Gawrelski (Wells Fargo): With a leading frontier model next year, where will value accrue—platforms or scaled first-party applications?
  Response: Value will accrue broadly across hardware, cloud and scaled applications, with outsized returns to firms that build best-in-class, widely scaled AI products; Meta emphasizes its strength in scaling apps to billions to capture that value.

Contradiction Point 1

AI Infrastructure Investments and Returns

It involves different perspectives on the expected returns from AI infrastructure investments, which are crucial for strategic planning and investor expectations.

Can you share early quantifiable signals from AB tests on upcoming improvements and confidence in ROIC from CapEx? How large is the Reality Labs revenue headwind in the 4Q guidance? - Eric Sheridan (Goldman Sachs)

2025Q3: We see strong ROI from core AI, and while genAI is earlier on the return curve, we're optimistic about monetization opportunities it will open up. - Susan Li(CFO)

What technological constraints or key factors are you focused on overcoming in the next 24 months to achieve leadership in superintelligence over the next decade? Additionally, what are the key factors you’re excited about that will drive engagement in the next 18 months? - Brian Nowak (Morgan Stanley)

2025Q2: CapEx will significantly increase in 2026 for genAI capacity, with a focus on core AI and infrastructure investments. - Susan Li(CFO)

Contradiction Point 2

AI Model Development and Integration

It highlights differing views on the timeline and approach to integrating AI models into existing products, impacting expectations for innovation and product development.

How does visibility into ad performance and engagement improvements compare to the past two years' progress? Should we expect an updated frontier model launch next year? - Mark Shmulik (Bernstein)

2025Q3: We expect both novel models and products, with no specific timing announced, but we are excited to share more when ready. - Mark Zuckerberg(CEO)

What key performance indicators (KPIs) does Meta track to measure progress in pursuing superintelligence, and how does the company balance revenue growth with investment pacing? - Mark Shmulik (Sanford C. Bernstein & Co., LLC)

2025Q2: We see significant potential in integrating foundation models like Behemoth across Meta's family of apps. Over the next 1 to 2 years, we expect to announce multiple AI-driven features. - Mark Zuckerberg(CEO)

Contradiction Point 3

Meta AI Adoption and Use

It highlights differing perspectives on Meta AI's adoption and usage, which are crucial for understanding the potential impact on the company's product strategy and user engagement.

What's encouraging about Meta AI adoption? Also, how far along are you in monetizing Meta AI? - Mark Mahaney (Evercore ISI)

2025Q3: We see increased usage as models improve, showing a strong runway for engagement. We're focusing on building leading models and then integrating them into products. - Mark Zuckerberg(CEO)

Could you highlight key advancements in Large Language Models (LLMs) and your focus areas? What is the global traction of Meta AI, and what recurring user behaviors have emerged from early use cases? - Brian Nowak (Morgan Stanley)

2025Q1: Meta AI sees strong engagement on WhatsApp and Facebook, with top use cases including information gathering, social interactions, and writing assistance. - Susan Li(CFO)

Contradiction Point 4

AI and CapEx Investments

It highlights differing perspectives on the strategic importance and expected outcomes of significant AI and CapEx investments, which are crucial for the company's future growth and technology leadership.

How are you approaching front-loading capacity for superintelligence and balancing CapEx growth with faster expense growth next year, and what impact will this have on earnings and free cash flow? Do you have targets for cash on hand or net cash? - Douglas Anmuth (JPMorgan)

2025Q3: We are working on our 2026 budget, focusing on ensuring we have the capacity needed for 2026, with flexibility for future needs. - Susan Li(CFO)

How will you monetize Meta AI? - Justin Post (Bank of America)

2024Q4: Our growth in 2026 CapEx relative to 2025 comes from growth in core areas, MSL, core AI as well as non-AI spend. - Susan Li(CFO)

Contradiction Point 5

AI and Pricing Dynamics

It involves differing explanations of the relationship between AI advancements and pricing dynamics, which could impact investor expectations regarding revenue growth and monetization strategies.

How is Advantage+ driving higher ROI for advertisers? - Ronald Josey (Citi)

2025Q3: Advantage+ leads to a 14% lower cost per lead for advertisers. - Susan Li(CFO)

How will pricing vs. impression growth impact future ad revenue? - Mark Shmulik (Bernstein)

2024Q4: Both pricing and impression growth contribute to ad revenue. Pricing benefits from improved ad performance and efficiency. - Susan Li(CFO)